A method and apparatus for adaptive filtering of phase sensitive optical time domain reflectometry signals

Abstract

This invention discloses an adaptive filtering method for phase-sensitive optical time-domain reflectometry (OTDR) signals. The method evaluates the quality of non-local mean filtering results by calculating correlation coefficients, and the filtering quality can be used to optimize the filtering parameters of the image filtering method. These parameters are first verified using a hypothetical standard signal and then applied to denoising a real signal from a phase-sensitive OTD instrument. Comparison of the correlation-optimized filtering parameters with those optimized using traditional peak signal-to-noise ratio (PSNR) and structural similarity methods shows that, except for slight differences in the neighborhood window size, the optimized filtering parameters are almost identical. This method is also applicable to image filtering of other two-dimensional periodic signals.

Description

Technical Field

[0001] This invention relates to the field of optical information sensing technology, and in particular to an adaptive filtering method and apparatus for phase-sensitive optical time-domain reflectance signals. Background Technology

[0002] Due to phase-sensitive optical time-domain reflectometry With advantages such as fast response speed, high sensitivity, and strong multiple detection capability, it has become increasingly popular in recent years. The research and application of phase-sensitive optical time-domain reflectometers (PTZAR) have received widespread attention, and they have been applied in fields such as perimeter intrusion detection, pipeline protection, and seismic mapping. By monitoring the Rayleigh backscattering statistical interference intensity of incident pulsed light, PTRAR systems can provide qualitative vibration measurements. However, coherent noise, white noise, and environmental interference can severely reduce the signal-to-noise ratio, degrading the detected signal and leading to a higher false alarm rate in intrusion detection systems or distortion due to frequency reconversion in acoustic sensing. Improving… The sensitivity of the system's detection has attracted researchers' interest and become a research focus in distributed fiber optic sensing. To improve the signal-to-noise ratio and compensate for fiber loss, many methods have been used. Long-range detection methods include pulse-width modulated Brillouin amplification, weak fiber Bragg grating arrays, and distributed Raman amplifiers. However, these methods rely on optical nonlinear effects, requiring considerable optical power and additional hardware costs. Another approach is through direct signal processing, which is the most effective method for noise reduction without increasing hardware costs. Different signal processing methods have been applied to analyze complex... Signal processing methods include moving average, moving differential, Fourier transform, wavelet denoising, mode decomposition, statistical clustering, empirical mode decomposition-Pearson correlation coefficient, and orthogonal matching pursuit. Recently, two-dimensional image restoration, the so-called nonlocal mean (NLM), has been shown to improve the signal-to-noise ratio of Brillouin and Raman scattering signals by up to 20 dB. This method greatly improves the signal by utilizing the information redundancy and correlation contained in multidimensional data, but it is time-consuming and unsuitable for applications requiring fast response times. The system. However, this drawback can be overcome by bilateral filtering, which can quickly detect [the problem] in a 27.6 km sensing fiber. The signal-to-noise ratio (SNR) was improved to 14 dB without spatial resolution loss. Regardless of the signal filtering method used, optimizing filter parameters is necessary to achieve optimal filtering quality, especially when comparing various filtering methods. Traditionally, evaluation is based on SNR, which typically requires a standard reference signal; however, obtaining a valid standard signal in the detection field is often difficult. Most current nonlocal mean filter parameter optimization methods are based on noise level estimation. These include methods such as calculating the standard deviation (STD) of measurements, filter-based estimation algorithms, and local block-based estimation algorithms that select weakly textured local blocks in the BOTDA signal and then use weakly textured principal component analysis (W-PCA) to estimate the noise level. These methods are complex and computationally inefficient. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide an adaptive filtering method and apparatus for phase-sensitive optical time-domain reflectance signals. Instead of using noise level estimation, the method utilizes the sensitivity of the autocorrelation function to the signal to optimize the filtering parameters of the non-local mean, evaluates the filtering quality of the two-dimensional image data through the correlation coefficient, and further evaluates the optimal filtering parameters, thus realizing adaptive filtering of phase-sensitive optical time-domain reflectance signals in a simple and convenient way.

[0004] In a first aspect, the present invention provides an adaptive filtering method for phase-sensitive optical time-domain reflectometry signals, applied to the vibration signal measurement of a phase-sensitive optical time-domain reflectometer, comprising the following steps:

[0005] Set the initial values ​​and step size of the filtering parameters for the nonlocal mean filter, including Gaussian parameters, neighborhood window parameters, and search window parameters;

[0006] The Gaussian parameters are adjusted from the initial value by a set step size, and the correlation coefficient is calculated at the same time. After each filtering, the peak of the first period of the correlation coefficient is recorded. When the peak of the first period of the autocorrelation coefficient reaches the maximum value, the corresponding Gaussian parameters are recorded as the optimal Gaussian parameters.

[0007] The Gaussian parameters are fixed as the optimal Gaussian parameters. Then, the neighborhood window parameters are adjusted from the initial value according to the set step size. At the same time, the peak of the first period of the correlation coefficient is calculated. When the peak of the first period of the correlation coefficient reaches the maximum value, the corresponding neighborhood window parameters are recorded as the optimal neighborhood window parameters.

[0008] Set the Gaussian parameter to the optimal Gaussian parameter and the neighborhood window parameter to the optimal neighborhood window parameter. Then, adjust the search window from the initial value according to the set step size. At the same time, calculate the first period peak of the correlation coefficient. When the first period peak of the correlation coefficient reaches the maximum value, record the corresponding search window parameter as the optimal search window parameter.

[0009] The correlation coefficient is calculated from the sequence correlation of the denoised image itself, utilizing the periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time-domain reflectometer.

[0010] Adaptive filtering of phase-sensitive optical temporal reflectance signals is performed using the optimal Gaussian parameters, optimal neighborhood window parameters, and optimal search window parameters.

[0011] Furthermore, the method for calculating the correlation coefficient specifically includes:

[0012] The denoising result of the target pixel is calculated as shown in formula (1):

[0013]

[0014] Where S is the neighboring window image block, t is the target window image block, I is the entire image, u(t) is the denoised target window image block, v(s) is the noisy neighboring window image block, and ω(s,t) is the weighting factor, which can be calculated by formula (2):

[0015]

[0016] Where h is the Gaussian smoothing parameter, v(t) is the image block to be denoised in the target window, and z(t) is the normalization coefficient, which can be calculated by formula (3):

[0017]

[0018] Autocorrelation, also known as sequence correlation, is the cross-correlation between a signal and itself at different points in time.

[0019]

[0020] Where R is a two-dimensional correlation coefficient matrix, and A is... The signal is the average value of matrix A, where m and n are the row and column of the matrix, respectively;

[0021] After obtaining the two-dimensional correlation coefficient matrix, the correlation coefficient along the trace direction is calculated as follows:

[0022]

[0023] Where i is the time delay of the signal sequence.

[0024] Secondly, the present invention provides an adaptive filtering device for phase-sensitive optical time-domain reflectometry signals, which is applied to the vibration signal measurement of a phase-sensitive optical time-domain reflectometer, including: an initialization module, a filtering parameter optimization module, and an adaptive filtering module;

[0025] The initialization module is used to set the initial values ​​and step size of the filtering parameters for nonlocal mean filtering. The filtering parameters include Gaussian parameters, neighborhood window parameters, and search window parameters.

[0026] The filter parameter optimization module is used to adjust the Gaussian parameter from the initial value according to a set step size, and at the same time calculate the first period peak of the correlation coefficient. When the first period peak of the correlation coefficient reaches the maximum value, the corresponding Gaussian parameter is recorded as the optimal Gaussian parameter.

[0027] The Gaussian parameters are fixed as the optimal Gaussian parameters. Then, the neighborhood window parameters are adjusted from the initial value according to the set step size. At the same time, the first period peak of the correlation coefficient is calculated. When the first period peak of the correlation coefficient reaches the maximum value, the corresponding neighborhood window parameters are recorded as the optimal neighborhood window parameters.

[0028] Set the Gaussian parameter to the optimal Gaussian parameter and the neighborhood window parameter to the optimal neighborhood window parameter. Then, adjust the search window from the initial value according to the set step size. At the same time, calculate the first period peak of the correlation coefficient. When the first period peak of the correlation coefficient reaches the maximum value, record the corresponding search window parameter as the optimal search window parameter.

[0029] The correlation coefficient is calculated from the sequence correlation of the denoised image itself, utilizing the periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time-domain reflectometer.

[0030] The adaptive filtering module is used to perform adaptive filtering of phase-sensitive light time-domain reflectance signals using the optimal Gaussian parameters, the optimal neighborhood window parameters, and the optimal search window parameters.

[0031] Furthermore, in the filter parameter optimization module, the method for calculating the correlation coefficient specifically includes:

[0032] The denoising result of the target pixel is calculated as shown in formula (1):

[0033]

[0034] Where S is the neighboring window image block, t is the target window image block, I is the entire image, u(t) is the denoised target window image block, v(s) is the noisy neighboring window image block, and ω(s,t) is the weighting factor, which can be calculated by formula (2):

[0035]

[0036] Where h is the Gaussian smoothing parameter, v(t) is the image block to be denoised in the target window, and z(t) is the normalization coefficient, which can be calculated by formula (3):

[0037]

[0038] Autocorrelation, also known as sequence correlation, is the cross-correlation between a signal and itself at different points in time.

[0039]

[0040] Where R is a two-dimensional correlation coefficient matrix, and A is... The signal is the average value of matrix A, where m and n are the row and column of the matrix, respectively;

[0041] After obtaining the two-dimensional correlation coefficient matrix, the correlation coefficient along the trace direction is calculated as follows:

[0042]

[0043] Where i is the time delay of the signal sequence.

[0044] Thirdly, the present invention provides an adaptive filtering method for phase-sensitive optical time-domain reflectometry signals, applied to the vibration signal measurement of a phase-sensitive optical time-domain reflectometer, comprising the following steps:

[0045] Set the initial values ​​and step size of the filtering parameters for the nonlocal mean filter, including Gaussian parameters, neighborhood window parameters, and search window parameters;

[0046] The Gaussian parameters are adjusted from the initial value by a set step size, and the correlation coefficient is calculated at the same time. After each filtering, the kurtosis or power density of the correlation coefficient is recorded. When the kurtosis or power density of the correlation coefficient reaches the minimum, the corresponding Gaussian parameters are recorded as the optimal Gaussian parameters.

[0047] The Gaussian parameters are fixed as the optimal Gaussian parameters. Then, the neighborhood window parameters are adjusted from the initial value according to the set step size. At the same time, the kurtosis or power density of the correlation coefficient is calculated. When the kurtosis or power density of the correlation coefficient reaches the minimum, the corresponding neighborhood window parameters are recorded as the optimal neighborhood window parameters.

[0048] Set the Gaussian parameter to the optimal Gaussian parameter and the neighborhood window parameter to the optimal neighborhood window parameter. Then, adjust the search window from the initial value according to the set step size. At the same time, calculate the kurtosis or power density of the correlation coefficient. When the kurtosis or power density of the correlation coefficient reaches the minimum, record the corresponding search window parameter as the optimal search window parameter.

[0049] The correlation coefficient is calculated from the sequence correlation of the denoised image itself, utilizing the periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time-domain reflectometer.

[0050] Adaptive filtering of phase-sensitive optical temporal reflectance signals is performed using the optimal Gaussian parameters, optimal neighborhood window parameters, and optimal search window parameters.

[0051] Fourthly, the present invention provides an adaptive filtering method for phase-sensitive optical time-domain reflectometry signals, applied to the vibration signal measurement of a phase-sensitive optical time-domain reflectometer, comprising the following steps:

[0052] Set the initial values ​​and step size of the filtering parameters for two-dimensional periodic signal image filtering;

[0053] The first filter parameter is adjusted from its initial value by a set step size, and the correlation coefficient is calculated. When the peak of the first period of the correlation coefficient reaches its maximum value, the value of the corresponding first filter parameter is recorded as the optimal first filter parameter.

[0054] The value of the first filter parameter is fixed as the optimal first filter parameter. Then the second filter parameter is adjusted from the initial value according to the set step size. At the same time, the correlation coefficient is calculated. When the peak of the first period of the correlation coefficient reaches the maximum value, the corresponding second filter parameter is recorded as the optimal second filter parameter.

[0055] Optimize all filter parameters using the same method;

[0056] The correlation coefficient is calculated from the sequence correlation of the denoised image itself, utilizing the periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time-domain reflectometer.

[0057] Adaptive filtering of phase-sensitive optical time-domain reflectance signals is performed using optimized filtering parameters.

[0058] Furthermore, the method for calculating the correlation coefficient specifically includes:

[0059] The denoising result of the target pixel is calculated as shown in formula (1):

[0060]

[0061] Where S is the neighboring window image block, t is the target window image block, I is the entire image, u(t) is the denoised target window image block, v(s) is the noisy neighboring window image block, and ω(s,t) is the weighting factor, which can be calculated by formula (2):

[0062]

[0063] Where h is the Gaussian smoothing parameter, v(t) is the image block to be denoised in the target window, and z(t) is the normalization coefficient, which can be calculated by formula (3):

[0064]

[0065] Autocorrelation, also known as sequence correlation, is the cross-correlation between a signal and itself at different points in time.

[0066]

[0067] Where R is a two-dimensional correlation coefficient matrix, and A is... The signal is the average value of matrix A, where m and n are the row and column of the matrix, respectively;

[0068] After obtaining the two-dimensional correlation coefficient matrix, the correlation coefficient along the trace direction is calculated as follows:

[0069]

[0070] Where i is the time delay of the signal sequence.

[0071] Fifthly, the present invention provides an adaptive filtering method for phase-sensitive optical time-domain reflectometry signals, applied to the vibration signal measurement of a phase-sensitive optical time-domain reflectometer, comprising the following steps:

[0072] Set the initial values ​​and step size of the filtering parameters for two-dimensional periodic signal image filtering;

[0073] The first filter parameter is adjusted from its initial value by a set step size. At the same time, the correlation coefficient is calculated. After each filtering, the kurtosis or power density of the correlation coefficient is recorded. When the kurtosis or power density of the correlation coefficient reaches its minimum, the value of the corresponding first filter parameter is recorded as the optimal first filter parameter.

[0074] The value of the first filter parameter is fixed as the optimal first filter parameter. Then, the second filter parameter is adjusted from the initial value according to the set step size. At the same time, the kurtosis or power density of the correlation coefficient is calculated. When the kurtosis or power density of the correlation coefficient reaches the minimum, the corresponding second filter parameter is recorded as the optimal second filter parameter.

[0075] Optimize all filter parameters using the same method;

[0076] The correlation coefficient is calculated from the sequence correlation of the denoised image itself, utilizing the periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time-domain reflectometer.

[0077] Adaptive filtering of phase-sensitive optical time-domain reflectance signals is performed using optimized filtering parameters.

[0078] The technical solutions provided in the embodiments of the present invention have the following technical effects or advantages:

[0079] 1. The optimal parameters are filtered by the correlation coefficient calculation results. The correlation coefficient method only relies on the autocorrelation of the vibration signal of the phase-sensitive optical time-domain reflectometer. It does not require a pure and noise-free reference signal. It can evaluate the filtering effect of the non-local mean filtering method and thus optimize the filtering parameters. Therefore, it is more suitable for field testing and has a wider range of applications.

[0080] 2. Compared with the traditional method of calculating the relationship between the optimal grayscale standard deviation and the noise standard deviation, and evaluating the filtering effect of the nonlocal mean filtering method based on the standard deviation calculation results, and then optimizing the filtering parameters, the algorithm of this invention only needs to calculate the autocorrelation coefficient of the image, which has a small computational load and high efficiency.

[0081] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0082] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0083] Figure 1 These are schematic diagrams of a standard image, a noisy image, and an image after NLM filtering in the prior art;

[0084] Figure 2 This is a schematic diagram illustrating the relationship and optimization of correlation coefficient, Gaussian parameter, target window size, and search window size when the present invention is verified using a assumed standard noisy image.

[0085] Figure 3 for A schematic diagram of the actual weak vibration signal of the meter-sensing optical fiber;

[0086] Figure 4 This is a schematic diagram illustrating the relationship and optimization of correlation coefficient, Gaussian parameter, target window size, and search window size when the present invention is tested using actual φ-OTDR signals.

[0087] Figure 5 This invention uses NLM denoising with optimized parameters. Schematic diagram of actual vibration signal;

[0088] Figure 6 This is a flowchart illustrating the method in Embodiment 1 of the present invention;

[0089] Figure 7 This is a schematic diagram of the device in Embodiment 2 of the present invention. Detailed Implementation

[0090] This application provides an adaptive filtering method and apparatus for phase-sensitive optical time-domain reflectometry signals. By using correlation coefficients to evaluate the filtering quality of two-dimensional image data, the optimal filtering parameters can be further evaluated, thus achieving adaptive filtering of phase-sensitive optical time-domain reflectometry signals in a simple and convenient manner.

[0091] The overall concept of the technical solution in this application is as follows:

[0092] Since noise itself is uncorrelated, and phase-sensitive optical time-domain reflectometers The vibration signal is periodic and strongly correlated. This invention proposes an adaptive filtering method for phase-sensitive optical time-domain reflectometry signals, which optimizes the filtering parameters based on adaptive NLM with correlation coefficients, thereby improving... Signal-to-noise ratio. This method is simple and effective; it can effectively evaluate the filtering quality of two-dimensional image data using only the correlation coefficient.

[0093] right The data pairs are digitally acquired, and the vibration curve is obtained by subtracting adjacent traces. Image denoising usually refers to average denoising, rather than machine learning-based denoising. The more average noise images, the better the denoising effect. By averaging all neighborhood blocks with similar gray levels and texture structures to denoise the target block, it was found that the more similar neighborhood image blocks, the better the denoising effect. Therefore, a larger search window centered on the target image block is needed. However, due to the limitations of computer computing power, the neighborhood size and the search window size should be appropriately set according to actual needs to balance the accuracy and efficiency of similarity calculation. The denoising result of the target pixel is the weighted average of all pixels, as shown in formula (1):

[0094]

[0095] Where S is the neighboring window image block, t is the target window image block, I is the entire image, u(t) is the denoised target window image block, v(s) is the noisy neighboring window image block, and ω(s,t) is the weighting factor, which can be calculated by formula (2):

[0096]

[0097] Where h is the Gaussian smoothing parameter, v(t) is the image block to be denoised in the target window, and z(t) is the normalization coefficient, which can be calculated by formula (3):

[0098]

[0099] Autocorrelation, also known as sequence correlation, is the cross-correlation between a signal and itself at different points in time.

[0100]

[0101] Where R is a two-dimensional correlation coefficient matrix, and A is... The signal is the average value of matrix A, where m and n are the row and column of the matrix, respectively;

[0102] After obtaining the two-dimensional correlation coefficient matrix, the correlation coefficient along the trace direction is calculated as follows:

[0103]

[0104] Where i is the time delay of the signal sequence.

[0105] In existing technologies, peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) are traditional signal measurement standards.

[0106] Peak signal-to-noise ratio (PSNR) is frequently used in the field of image compression to measure the quality of signal reconstruction.

[0107]

[0108] Here, MSE is the mean square error between the original image (standard signal) and the denoised image. The larger the PSNR value, the less distortion of the denoised signal. Therefore, the larger the PSNR, the better the filtering effect of the image.

[0109] Structural similarity (SSIM) is a visual metric that measures the similarity between two images (a filtered image and a standard signal). Let x be the original image and y be the denoised image.

[0110]

[0111] μ x and μ y It is the average brightness of the image. and It is the standard deviation, σ xy This represents the covariance, where c1, c2, and c3 are constants. The larger this value, the better the SSIM of y and x.

[0112] If the PSNR and SSIM values ​​are large, it indicates that the filtered image is close to the original image signal, resulting in better visual quality. However, both PSNR and SSIM calculations require a standard reference signal, which is often difficult to obtain at the detection site.

[0113] Since two-dimensional image restoration, i.e., nonlocal mean (NLM), has been shown to improve the signal-to-noise ratio of Brillouin and Raman scattering signals by up to 20 dB, this embodiment of the invention forms an adaptive NLM based on the calculation of the correlation coefficient, optimizes the filtering parameters to obtain the best filtering quality, thereby improving... The signal-to-noise ratio (SNR) of the signal. To verify the effectiveness of the method in this embodiment of the invention, it is first verified using a hypothetical standard noise image, and then verified using actual... The signal is tested.

[0114] First, to understand the relationship between correlation coefficient, PSNR, and SSIM, the following experiment was conducted. First, a noise-free standard signal was generated, such as... Figure 1 As shown in (a), random noise of different amplitude levels from 0.2 to 1 is added to the standard signal image, as follows: Figure 1 As shown in (b). Figure 1 (c) is the image after NLM filtering. The original image size is 126×126. Next, to find the optimal Gaussian filtering parameters, with the neighborhood and search window sizes set to 10 and 100 respectively, the noisy image was denoised using Gaussian parameters increasing from 0.02 to 0.12 in steps of 0.004. Figure 2 As shown in (a), the Gaussian parameters are optimal when the correlation coefficient of the first period (asterisk) of the NLM-filtered image reaches its maximum. Traditionally, the Gaussian filter parameters are optimal when the PSNR and SSIM of the image are maximized. This experiment compares the optimal Gaussian filter parameters obtained by the three methods with the increase in noise amplitude, such as... Figure 2 As shown in (b), they increase with increasing noise amplitude. At the same noise level, the optimal Gaussian parameters obtained by these methods are almost identical.

[0115] Furthermore, the same process is used to find the optimal neighborhood window size and search window size, such as... Figure 2 (c) and Figure 2 As shown in (d), the optimal values ​​for the neighborhood window and search window remain essentially unchanged as the noise amplitude increases. The optimal neighborhood window sizes obtained from the correlation coefficient, PSNR, and SSIM are 14, 8, and 11, respectively. It can be seen from the figure that the neighborhood window size obtained from the correlation coefficient is close to half the signal period, while the one obtained from PSNR is about one-quarter of the signal period. The result from SSIM falls between the two methods. For the search window, except for larger bias data, the optimal search window size obtained using correlation, PSNR, and SSIM is 251.

[0116] Then, to verify the effectiveness of the adaptive NLM filter, the relevant algorithm was applied... The actual signal processing involves subtracting 101 adjacent trajectories to obtain the signal image, as shown below. Figure 3 As shown. The autocorrelation function of a periodic signal is still that of a periodic signal, such as... Figure 4As shown in (a). If the time delay of a periodic signal reaches one period, the autocorrelation function of the signal will reach its peak (asterisk). The asterisk indicates that the correlation coefficient reaches its maximum value when the offset is 9. Since noise is not correlated, filtering only noise will not change the correlation coefficient, but removing the effective components of the signal will reduce the correlation coefficient. Therefore, the correlation coefficient can be used to evaluate the filtering quality and optimize the filtering parameters.

[0117] The Gaussian parameter controls the degree of attenuation of the Gaussian function. A larger Gaussian function results in a smoother transition and a higher level of noise reduction, but it can also lead to a more blurred image. Conversely, more edge details are retained, but more noise is also preserved. When the Gaussian parameter changes from 0.033 to 0.058, the step size is 0.002, and the neighborhood and search window sizes are set to 10 and 80 respectively, the relationship between the correlation parameters of the filtered signal and the Gaussian parameter and correlation coefficient is as follows: Figure 4 As shown in (b), the correlation coefficient increases with the increase of the Gaussian smoothing parameter due to the reduction of noise, and then decreases due to the filtering out of some effective components of the signal. It reaches a peak of 0.3716 when the Gaussian parameter is optimal at 0.047.

[0118] The neighborhood window in NLM preserves detailed information when filtering images. If the neighborhood window is only one pixel in size, NLM can be considered a bilateral filtering method. The relationship between the correlation coefficient and the neighborhood window size is as follows when the smoothing parameter and search window are set to 0.047 and 80, respectively. Figure 4 As shown in (c), the correlation coefficient initially increases with the increase of the neighborhood window size. When the neighborhood window size is 9, the correlation coefficient reaches its maximum value of 0.4219, and then decreases slightly. Because all pixels in the neighborhood window participate in the similarity calculation, a large neighborhood window leads to a large computational burden. Therefore, the optimal neighborhood window not only improves the filtering quality but also saves computational resources. The search window of NLM searches for similar neighborhood image patches to denoise the target image patch. In principle, the larger the search window, the more neighborhood windows can be found, and the better the filtering effect. Based on the previously determined optimal Gaussian parameter of 0.047 and a neighborhood window size of 9, the relationship between the correlation coefficient and the search window size is plotted on... Figure 4 In (d), the correlation coefficient initially increases and then decreases rapidly as the search window size increases. The correlation coefficient reaches its maximum of 0.4453 when the search window size is 196. The optimal search window size is approximately twice the dimension of the data matrix. Therefore, the search window size should be as large as possible, provided that computer performance allows. In practice, the search window size should be a trade-off between filtering effectiveness and computational cost.

[0119] After optimizing the three filtering parameters of NLM, it can be seen that the autocorrelation coefficient gradually increases from 0.3716 to 0.4302. Ultimately, The NLM filtering result of the signal is as follows Figure 5 As shown, it clearly demonstrates that under the influence of a light breeze, the first 40 meters of the sensing fiber suspended in the air vibrate slightly, while the subsequent 60 meters are almost unaffected. This result is in excellent agreement with the effects of a light breeze on the sensing fiber. Figure 3 Compared to the original signal, this adaptive NLM filtering greatly improves image quality.

[0120] The optimized filtering parameters were compared with those optimized using traditional peak signal-to-noise ratio and structural similarity algorithms. The results showed that the optimized filtering parameters were almost identical, except for a slight difference in the neighborhood window size. This method is also applicable to filtering other two-dimensional periodic signal images.

[0121] Example 1

[0122] This embodiment provides an adaptive filtering method for phase-sensitive optical time-domain reflectometry signals, applied to vibration signal measurement of a phase-sensitive optical time-domain reflectometer, such as... Figure 6 As shown, it includes:

[0123] Set the initial values ​​and step size of the filtering parameters for the nonlocal mean filter, including Gaussian parameters, neighborhood window parameters, and search window parameters;

[0124] The Gaussian parameters are adjusted from the initial value by a set step size, and the correlation coefficient is calculated. When the correlation coefficient reaches the first peak value, the corresponding Gaussian parameters are recorded as the optimal Gaussian parameters.

[0125] The Gaussian parameters are fixed as the optimal Gaussian parameters. Then, the neighborhood window parameters are adjusted from the initial value according to the set step size. At the same time, the correlation coefficient is calculated. When the correlation coefficient reaches the second peak, the corresponding neighborhood window parameters are recorded as the optimal neighborhood window parameters.

[0126] Set the Gaussian parameter to the optimal Gaussian parameter and the neighborhood window parameter to the optimal neighborhood window parameter. Then, adjust the search window from the initial value according to the set step size, and calculate the correlation coefficient. When the correlation coefficient reaches the third peak, record the corresponding search window parameter as the optimal search window parameter.

[0127] The correlation coefficient is calculated by utilizing the periodicity and autocorrelation characteristics of the vibration signal from a phase-sensitive optical time-domain reflectometer, and is obtained from the sequence correlation between the denoised image and the original noisy image.

[0128] Adaptive filtering of phase-sensitive optical temporal reflectance signals is performed using the optimal Gaussian parameters, optimal neighborhood window parameters, and optimal search window parameters.

[0129] Preferably, the method for calculating the correlation coefficient specifically includes:

[0130] The denoising result of the target pixel is calculated as shown in formula (1):

[0131]

[0132] Where S is the neighboring window image block, t is the target window image block, I is the entire image, u(t) is the denoised target window image block, v(s) is the noisy neighboring window image block, and ω(s,t) is the weighting factor, which can be calculated by formula (2):

[0133]

[0134] Where h is the Gaussian smoothing parameter, v(t) is the image block to be denoised in the target window, and z(t) is the normalization coefficient, which can be calculated by formula (3):

[0135]

[0136] Autocorrelation, also known as sequence correlation, is the cross-correlation between a signal and itself at different points in time.

[0137]

[0138] Where R is a two-dimensional correlation coefficient matrix, and A is... The signal is the average value of matrix A, where m and n are the row and column of the matrix, respectively;

[0139] The correlation coefficient along the trajectory, i.e., the time delay, is calculated using the following formula:

[0140]

[0141] Where i is the time delay of the signal sequence.

[0142] Based on the same inventive concept, this application also provides an apparatus corresponding to the method in Embodiment 1, as detailed in Embodiment 2.

[0143] Example 2

[0144] This embodiment provides a phase-sensitive optical time-domain reflectometry (OTDR) adaptive filtering device, applied to vibration signal measurement of a phase-sensitive OTD meter, such as... Figure 7 As shown, it includes: an initialization module, a filter parameter optimization module, and an adaptive filtering module;

[0145] The initialization module is used to set the initial values ​​and step size of the filtering parameters for nonlocal mean filtering. The filtering parameters include Gaussian parameters, neighborhood window parameters, and search window parameters.

[0146] The filter parameter optimization module is used to adjust the Gaussian parameter from the initial value according to a set step size, and at the same time calculate the correlation coefficient. When the correlation coefficient reaches the first peak value, the corresponding Gaussian parameter is recorded as the optimal Gaussian parameter.

[0147] The Gaussian parameters are fixed as the optimal Gaussian parameters. Then, the neighborhood window parameters are adjusted from the initial value according to the set step size. At the same time, the correlation coefficient is calculated. When the correlation coefficient reaches the second peak, the corresponding neighborhood window parameters are recorded as the optimal neighborhood window parameters.

[0148] Set the Gaussian parameter to the optimal Gaussian parameter and the neighborhood window parameter to the optimal neighborhood window parameter. Then, adjust the search window from the initial value according to the set step size, and calculate the correlation coefficient. When the correlation coefficient reaches the third peak, record the corresponding search window parameter as the optimal search window parameter.

[0149] The correlation coefficient is calculated by utilizing the periodicity and autocorrelation characteristics of the vibration signal from a phase-sensitive optical time-domain reflectometer, and is obtained from the sequence correlation between the denoised image and the original noisy image.

[0150] The adaptive filtering module is used to perform adaptive filtering of phase-sensitive light time-domain reflectance signals using the optimal Gaussian parameters, the optimal neighborhood window parameters, and the optimal search window parameters.

[0151] Preferably, in the filter parameter optimization module, the method for calculating the correlation coefficient specifically includes:

[0152] The denoising result of the target pixel is calculated as shown in formula (1):

[0153]

[0154] Where S is the neighboring window image block, t is the target window image block, I is the entire image, u(t) is the denoised target window image block, v(s) is the noisy neighboring window image block, and ω(s,t) is the weighting factor, which can be calculated by formula (2):

[0155]

[0156] Where h is the Gaussian smoothing parameter, v(t) is the image block to be denoised in the target window, and z(t) is the normalization coefficient, which can be calculated by formula (3):

[0157]

[0158] Autocorrelation, also known as sequence correlation, is the cross-correlation between a signal and itself at different points in time.

[0159]

[0160] Where R is a two-dimensional correlation coefficient matrix, and A is... The signal is the average value of matrix A, where m and n are the row and column of the matrix, respectively;

[0161] The correlation coefficient along the trajectory, i.e., the time delay, is calculated using the following formula:

[0162]

[0163] Where i is the time delay of the signal sequence.

[0164] Since the apparatus described in Embodiment 2 of the present invention is an apparatus used to implement the method of Embodiment 1 of the present invention, those skilled in the art can understand the specific structure and variations of the apparatus based on the method described in Embodiment 1 of the present invention, and therefore will not be described again here. All apparatuses used in the method of Embodiment 1 of the present invention fall within the scope of protection of the present invention.

[0165] Example 3

[0166] This embodiment provides an adaptive filtering method for phase-sensitive optical time-domain reflectometry signals, applied to vibration signal measurement of a phase-sensitive optical time-domain reflectometer. Unlike Embodiment 1, this embodiment uses the minimum value of kurtosis or power density as the criterion for determining the optimal parameters, and may include the following steps:

[0167] Set the initial values ​​and step size of the filtering parameters for the nonlocal mean filter, including Gaussian parameters, neighborhood window parameters, and search window parameters;

[0168] The Gaussian parameters are adjusted from the initial value by a set step size, and the correlation coefficient is calculated at the same time. After each filtering, the kurtosis or power density of the correlation coefficient is recorded. When the kurtosis or power density of the correlation coefficient reaches the minimum, the corresponding Gaussian parameters are recorded as the optimal Gaussian parameters.

[0169] The Gaussian parameters are fixed as the optimal Gaussian parameters. Then, the neighborhood window parameters are adjusted from the initial value according to the set step size. At the same time, the kurtosis or power density of the correlation coefficient is calculated. When the kurtosis or power density of the correlation coefficient reaches the minimum, the corresponding neighborhood window parameters are recorded as the optimal neighborhood window parameters.

[0170] Set the Gaussian parameter to the optimal Gaussian parameter and the neighborhood window parameter to the optimal neighborhood window parameter. Then, adjust the search window from the initial value according to the set step size. At the same time, calculate the kurtosis or power density of the correlation coefficient. When the kurtosis or power density of the correlation coefficient reaches the minimum, record the corresponding search window parameter as the optimal search window parameter.

[0171] The correlation coefficient is calculated from the sequence correlation of the denoised image itself, utilizing the periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time-domain reflectometer.

[0172] Adaptive filtering of phase-sensitive optical temporal reflectance signals is performed using the optimal Gaussian parameters, optimal neighborhood window parameters, and optimal search window parameters.

[0173] The method for calculating the correlation coefficient can be the same as in the previous embodiments.

[0174] Example 4

[0175] Besides local mean filtering, the above method can also be applied to other two-dimensional periodic signal image filtering. Therefore, this embodiment provides an adaptive filtering method for phase-sensitive optical time-domain reflectometry signals, applied to vibration signal measurement of a phase-sensitive optical time-domain reflectometer, including the following steps:

[0176] Set the initial values ​​and step size of the filtering parameters for two-dimensional periodic signal image filtering;

[0177] The first filter parameter is adjusted from its initial value by a set step size, and the correlation coefficient is calculated. When the peak of the first period of the correlation coefficient reaches its maximum value, the value of the corresponding first filter parameter is recorded as the optimal first filter parameter.

[0178] The value of the first filter parameter is fixed as the optimal first filter parameter. Then the second filter parameter is adjusted from the initial value according to the set step size. At the same time, the correlation coefficient is calculated. When the peak of the first period of the correlation coefficient reaches the maximum value, the corresponding second filter parameter is recorded as the optimal second filter parameter.

[0179] Optimize all filter parameters using the same method;

[0180] The correlation coefficient is calculated from the sequence correlation of the denoised image itself, utilizing the periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time-domain reflectometer.

[0181] Adaptive filtering of phase-sensitive optical time-domain reflectance signals is performed using optimized filtering parameters.

[0182] The method of optimizing filter parameters through relevant algorithms can also be applied to filtering other two-dimensional periodic signal images. It is only necessary to optimize the filter parameters sequentially and find the optimal filter parameters based on the value of the peak of the first period of the correlation coefficient to achieve adaptive filtering of the time-domain reflectance signal of the row phase-sensitive light.

[0183] Preferably, the method for calculating the correlation coefficient specifically includes:

[0184] The denoising result of the target pixel is calculated as shown in formula (1):

[0185]

[0186] Where S is the neighboring window image block, t is the target window image block, I is the entire image, u(t) is the denoised target window image block, v(s) is the noisy neighboring window image block, and ω(s,t) is the weighting factor, which can be calculated by formula (2):

[0187]

[0188] Where h is the Gaussian smoothing parameter, v(t) is the image block to be denoised in the target window, and z(t) is the normalization coefficient, which can be calculated by formula (3):

[0189]

[0190] Autocorrelation, also known as sequence correlation, is the cross-correlation between a signal and itself at different points in time.

[0191]

[0192] Where R is a two-dimensional correlation coefficient matrix, and A is... The signal is the average value of matrix A, where m and n are the row and column of the matrix, respectively;

[0193] The correlation coefficient along the trajectory, i.e., the time delay, is calculated using the following formula:

[0194]

[0195] Where i is the time delay of the signal sequence.

[0196] Example 5

[0197] Besides local mean filtering, the above method can also be applied to other two-dimensional periodic signal image filtering. Therefore, this embodiment provides an adaptive filtering method for phase-sensitive optical time-domain reflectometry signals, applied to vibration signal measurement of a phase-sensitive optical time-domain reflectometer. The difference from Embodiment 4 is that this embodiment uses the observation of whether the kurtosis or power density value reaches its minimum as the criterion for judging the optimal parameter, and may include the following steps:

[0198] Set the initial values ​​and step size of the filtering parameters for two-dimensional periodic signal image filtering;

[0199] The first filter parameter is adjusted from its initial value by a set step size. At the same time, the correlation coefficient is calculated. After each filtering, the kurtosis or power density of the correlation coefficient is recorded. When the kurtosis or power density of the correlation coefficient reaches its minimum, the value of the corresponding first filter parameter is recorded as the optimal first filter parameter.

[0200] The value of the first filter parameter is fixed as the optimal first filter parameter. Then, the second filter parameter is adjusted from the initial value according to the set step size. At the same time, the kurtosis or power density of the correlation coefficient is calculated. When the kurtosis or power density of the correlation coefficient reaches the minimum, the corresponding second filter parameter is recorded as the optimal second filter parameter.

[0201] Optimize all filter parameters using the same method;

[0202] The correlation coefficient is calculated from the sequence correlation of the denoised image itself, utilizing the periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time-domain reflectometer.

[0203] Adaptive filtering of phase-sensitive optical time-domain reflectance signals is performed using optimized filtering parameters.

[0204] The method for calculating the correlation coefficient can be the same as in the previous embodiments.

[0205] This invention filters optimal parameters by calculating the correlation coefficient. The correlation coefficient method relies solely on the autocorrelation of the vibration signal from a phase-sensitive optical time-domain reflectometer, without requiring a clean, noise-free reference signal. It can evaluate the filtering effect of nonlocal mean filtering methods and thus optimize filtering parameters. Therefore, it is more suitable for on-site testing and has a wider range of applications. Compared to the traditional method of calculating the relationship between the optimal grayscale standard deviation and the noise standard deviation of an image and then evaluating the filtering effect of nonlocal mean filtering methods to optimize filtering parameters, the algorithm of this invention only needs to calculate the autocorrelation coefficient of the image, resulting in low computational load and high efficiency.

[0206] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method of adaptive filtering of phase sensitive optical time domain reflectometry signals, characterized by, The application is applied to vibration signal measurement of phase-sensitive optical time domain reflectometer, and comprises the following steps: initial values and steps of filter parameters of non-local mean filtering are set, the filter parameters comprise Gaussian parameters, neighborhood window parameters and search window parameters; the Gaussian parameters are adjusted from the initial values by the set steps, and the correlation coefficient is calculated, after each filtering, the first period wave crest of the correlation coefficient is recorded, when the first period wave crest of the correlation coefficient reaches the maximum value, the corresponding Gaussian parameter is recorded as the optimal Gaussian parameter; the Gaussian parameters are fixed as the optimal Gaussian parameters, then the neighborhood window parameters are adjusted from the initial values by the set steps, and the first period wave crest of the correlation coefficient is calculated, when the first period wave crest of the correlation coefficient reaches the maximum value, the corresponding neighborhood window parameter is recorded as the optimal neighborhood window parameter; the Gaussian parameters are set as the optimal Gaussian parameters, the neighborhood window parameters are set as the optimal neighborhood window parameters, then the search window is adjusted from the initial value by the set step, and the first period wave crest of the correlation coefficient is calculated, when the first period wave crest of the correlation coefficient reaches the maximum value, the corresponding search window parameter is recorded as the optimal search window parameter; the correlation coefficient is calculated from the sequence correlation of the denoised image itself by using the periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time domain reflectometer; optimal Gaussian parameters, optimal neighborhood window parameters and optimal search window parameters are used for adaptive filtering of the phase-sensitive optical time domain reflector signal.

2. The method of claim 1, wherein: The calculation method of the correlation coefficient comprises: the denoising result of the target pixel is calculated, as shown in formula (1): (1) wherein S is a neighborhood window image block, t is a target window image block, I is the entire image, u(t) is a target window denoised image block, v(s) is a neighborhood window noise image block, ω(s, t) is a weighting factor, which can be calculated by equation (2): (2) wherein h is a Gaussian smoothing parameter, v(t) is the image block to be denoised in the target window; z(t) is a normalization coefficient, which can be calculated by equation (3): (3) The autocorrelation is also called sequence correlation, which is the cross-correlation between the signal and itself at different time points; (4) wherein R is a two-dimensional correlation coefficient matrix, A is φ - the signal of the OTDR, is the matrix A the average value of the matrix m and n are the rows and columns of the matrix, respectively; After the two-dimensional correlation coefficient matrix is obtained, the correlation coefficient of the correlation coefficient matrix along the trace direction is calculated, as follows: (5) wherein i is the time delay of the signal sequence.

3. A phase sensitive optical time domain reflectometry signal adaptive filtering apparatus, characterized by, The application is applied to vibration signal measurement of phase-sensitive optical time domain reflectometer, and comprises: an initialization module, a filter parameter optimization module and an adaptive filtering module; The initialization module is used for setting initial values and steps of filter parameters of non-local mean filtering, the filter parameters comprise Gaussian parameters, neighborhood window parameters and search window parameters; The filter parameter optimization module is used for adjusting the Gaussian parameters from the initial values by the set steps, and calculating the first period wave crest of the correlation coefficient, when the first period wave crest of the correlation coefficient reaches the maximum value, the corresponding Gaussian parameter is recorded as the optimal Gaussian parameter; the Gaussian parameters are fixed as the optimal Gaussian parameters, then the neighborhood window parameters are adjusted from the initial values by the set steps, and the first period wave crest of the correlation coefficient is calculated, when the first period wave crest of the correlation coefficient reaches the maximum value, the corresponding neighborhood window parameter is recorded as the optimal neighborhood window parameter; The Gaussian parameter is set as the optimal Gaussian parameter, the neighborhood window parameter is set as the optimal neighborhood window parameter, and then the search window is adjusted from an initial value by a set step size, while the first period wave crest of the correlation coefficient is calculated, when the first period wave crest of the correlation coefficient reaches a maximum value, the corresponding search window parameter is recorded as the optimal search window parameter; The correlation coefficient is calculated by using periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time domain reflectometer, and by using sequence correlation of the denoised image itself; The adaptive filtering module is configured to perform adaptive filtering on the phase-sensitive optical time domain reflection signal by using the optimal Gaussian parameter, the optimal neighborhood window parameter, and the optimal search window parameter.

4. The apparatus of claim 3, wherein: In the filtering parameter optimization module, the method for calculating the correlation coefficient comprises: The denoising result of the target pixel is calculated, as shown in formula (1): (1) wherein S is a neighborhood window image block, t is a target window image block, I is the entire image, u(t) is a target window denoised image block, v(s) is a neighborhood window noise image block, ω(s, t) is a weighting factor, which can be calculated by equation (2): (2) wherein h is a Gaussian smoothing parameter, v(t) is the image block to be denoised in the target window; z(t) is a normalization coefficient, which can be calculated by equation (3): (3) Autocorrelation is also called sequence correlation, which is the cross-correlation between a signal and itself at different time points; (4) wherein R is a two-dimensional correlation coefficient matrix, A is φ - the signal of the OTDR, is the matrix A the average value of the matrix m and n are the rows and columns of the matrix, respectively; After obtaining the two-dimensional correlation coefficient matrix, the correlation coefficient in the trace direction is calculated, as shown in the following formula: (5) wherein i is the time delay of the signal sequence.

5. A method of adaptive filtering of phase sensitive optical time domain reflectometry signals, characterized by, The application is applied to vibration signal measurement of a phase-sensitive optical time domain reflectometer, and comprises the following steps: Initial values and step sizes of filtering parameters of the non-local mean filtering are set, the filtering parameters including a Gaussian parameter, a neighborhood window parameter, and a search window parameter; The Gaussian parameter is adjusted from an initial value by a set step size, while the correlation coefficient is calculated, after each filtering, the kurtosis or power density of the correlation coefficient is recorded, when the kurtosis or power density of the correlation coefficient reaches a minimum value, the corresponding Gaussian parameter is recorded as the optimal Gaussian parameter; The Gaussian parameter is fixed as the optimal Gaussian parameter, and then the neighborhood window parameter is adjusted from an initial value by a set step size, while the kurtosis or power density of the correlation coefficient is calculated, when the kurtosis or power density of the correlation coefficient reaches a minimum value, the corresponding neighborhood window parameter is recorded as the optimal neighborhood window parameter; The Gaussian parameter is set as the optimal Gaussian parameter, the neighborhood window parameter is set as the optimal neighborhood window parameter, and then the search window is adjusted from an initial value by a set step size, while the kurtosis or power density of the correlation coefficient is calculated, when the kurtosis or power density of the correlation coefficient reaches a minimum value, the corresponding search window parameter is recorded as the optimal search window parameter; The correlation coefficient is calculated by using periodicity and autocorrelation characteristics of the vibration signal of the phase-sensitive optical time domain reflectometer, and by using sequence correlation of the denoised image itself; The adaptive filtering module is configured to perform adaptive filtering on the phase-sensitive optical time domain reflection signal by using the optimal Gaussian parameter, the optimal neighborhood window parameter, and the optimal search window parameter.

Images